entropy calculation
amit
the trajectory generated by
namd simulations. I have few questions:
(i) In the end of the calculation I get the following
"The Andricioaei entropy can not be calculated due to the presence of
an eigenvalue less than or equal to zero
Entropy (Schlitter) is 1792.364959 (J/molK)"
done.
Here, I have performed the calculations merging the trajectories of
(30ns +30ns + 30ns) together.
However when i perform the calculations on single trajectories (only
30ns) i get the value of Andricioaei entropy.
Do you think there is problem when i merge the two trajectories?
(ii) I have also observed the calculations is very dependent on the
"step" on specified in the command:
carma64 -v -cov -eigen -mass -temp 310 -atmid ALLID -segname A -step
1000 -last 45000 carma.fitted.dcd prot_sel.psf
As i get different values for when i put the value for instance to 100
or when 1000.
(iii) Is is possible to estimate also the error in the entropy
calculations?
(iv) is it possible to have a data file with the values of entropy as
a function of time?
many thanks in advance
best regards,
Amit
Nikolaos Glykos
Hi Amit,
> (i) In the end of the calculation I get the following
>
> "The Andricioaei entropy can not be calculated due to the presence of
> an eigenvalue less than or equal to zero
> Entropy (Schlitter) is 1792.364959 (J/molK)"
> done.
>
> Here, I have performed the calculations merging the trajectories of
> (30ns +30ns + 30ns) together.
> However when i perform the calculations on single trajectories (only
> 30ns) i get the value of Andricioaei entropy.
>
> Do you think there is problem when i merge the two trajectories?
Have you removed overall translations/rotations in the merged
trajectory ?
If the three runs are completely independent (and not restarts), I have
doubts
whether the calculation you describe is meaningful (the merged
trajectory will
have discontinuities, which will be reproduced from the principal
components
which I think invalidates the procedure implemented by carma).
> (ii) I have also observed the calculations is very dependent on the
> "step" on specified in the command:
>
> carma64 -v -cov -eigen -mass -temp 310 -atmid ALLID -segname A -step
> 1000 -last 45000 carma.fitted.dcd prot_sel.psf
>
> As i get different values for when i put the value for instance to
> 100
> or when 1000.
This is not unexpected. The variance-covariance matrix, and thus the
eigenvalues,
and thus the entropy estimate, will be very different if you use 1,000
frames instead
of 1,000,000 frames.
> (iii) Is is possible to estimate also the error in the entropy
> calculations?
The quick answer is that I do not know. If, however, you had a
sufficiently
sampled long trajectory, then you could repeat the calculations with
two
independent halves and then examine the deviation on your entropy
estimates.
Just a thought.
> (iv) is it possible to have a data file with the values of entropy as
> a function of time?
Not automagically.
--
Dr Nicholas M. Glykos, Department of Molecular Biology
and Genetics, Democritus University of Thrace, University Campus,
Dragana, 68100 Alexandroupolis, Greece, Tel/Fax (office)
+302551030620,
Ext.77620, Tel (lab) +302551030615, http://utopia.duth.gr/~glykos/
Michel Espinoza-Fonseca
As for your last question, I assume you want to evaluate the convergence of entropy calculation. You can simply calculate the entropy at times t1, t2...tn and save the entropy values in a separate file.
Best,
Michel
________________________________________
From: carma-molecu...@googlegroups.com [carma-molecu...@googlegroups.com] On Behalf Of Nikolaos Glykos [gly...@mbg.duth.gr]
Sent: Tuesday, September 27, 2011 11:24 AM
To: amit
Cc: carma molecular dynamics
Subject: Re: [carma] entropy calculation
Hi Amit,
Not automagically.
--
--
You received this message because you are subscribed to the Google Groups "carma molecular dynamics" group.
To post to this group, send email to carma-molecu...@googlegroups.com.
To unsubscribe from this group, send email to carma-molecular-dy...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/carma-molecular-dynamics?hl=en.
amit
thanks for your reply.
> Have you removed overall translations/rotations in the merged
> trajectory ?
> If the three runs are completely independent (and not restarts), I have
> doubts
> whether the calculation you describe is meaningful (the merged
> trajectory will
> have discontinuities, which will be reproduced from the principal
> components
> which I think invalidates the procedure implemented by carma).
>
merged trajectory.
They are restarts from the previous trajectory, but well written
without any errors.
Here, I am evaluating the entropy for a few selected residues in the
binding site.
I did not understand why you say that my calculations are not
meaningful. to do prefer
to the value of the entropy estimated not being meaningful?
> > (ii) I have also observed the calculations is very dependent on the
> > "step" on specified in the command:
>
> > carma64 -v -cov -eigen -mass -temp 310 -atmid ALLID -segname A -step
> > 1000 -last 45000 carma.fitted.dcd prot_sel.psf
>
> > As i get different values for when i put the value for instance to
> > 100
> > or when 1000.
>
> This is not unexpected. The variance-covariance matrix, and thus the
> eigenvalues,
> and thus the entropy estimate, will be very different if you use 1,000
> frames instead
> of 1,000,000 frames.
>
simulations
its advisable to have same number of frames?
> > (iii) Is is possible to estimate also the error in the entropy
> > calculations?
>
> The quick answer is that I do not know. If, however, you had a
> sufficiently
> sampled long trajectory, then you could repeat the calculations with
> two
> independent halves and then examine the deviation on your entropy
> estimates.
> Just a thought.
>
> > (iv) is it possible to have a data file with the values of entropy as
> > a function of time?
>
> Not automagically.
>
> --
>
> Dr Nicholas M. Glykos, Department of Molecular Biology
> and Genetics, Democritus University of Thrace, University Campus,
> Dragana, 68100 Alexandroupolis, Greece, Tel/Fax (office)
> +302551030620,
> Ext.77620, Tel (lab) +302551030615,http://utopia.duth.gr/~glykos/
Amit
amit
Yes, i wanted to evaluate the convergence of the entropy calculation.
I i understood well you reply for example if my simulation is for
100ns, then
i calculate the entropy for 10, 20, 30, 40, ...., 100 ns and look at
the values?
thanks
Amit
On Sep 27, 6:29 pm, Michel Espinoza-Fonseca <m...@ddt.biochem.umn.edu>
wrote:
Michel Espinoza-Fonseca
0-20...0-100 intervals.
Best,
Michel
On Sep 27, 2011, at 7:53 PM, "amit" <ami...@gmail.com> wrote:
> Hi Michel,
>
> Yes, i wanted to evaluate the convergence of the entropy calculation.
> I i understood well you reply for example if my simulation is for
> 100ns, then
> i calculate the entropy for 10, 20, 30, 40, ...., 100 ns and look at
> the values?
>
> thanks
> Amit
>
> On Sep 27, 6:29 pm, Michel Espinoza-Fonseca <m...@ddt.biochem.umn.edu>
> wrote:
>> Hi Amit,
>>
>> As for your last question, I assume you want to evaluate the
>> convergence of entropy calculation. You can simply calculate the
>> entropy at times t1, t2...tn and save the entropy values in a
>> separate file.
>>
>> Best,
>> Michel
>> ________________________________________
>> From: carma-molecu...@googlegroups.com [carma-molecular-
>> dyna...@googlegroups.com] On Behalf Of Nikolaos Glykos
Nicholas M Glykos
> yes, I have removed the overall translation and rotations in the merged
> trajectory. They are restarts from the previous trajectory, but well
> written without any errors.
You are ok then. I was refering to the possibility of completely
independent trajectories. Since they are restarts you should be fine.
> Here, I am evaluating the entropy for a few selected residues in the
> binding site.
The negative eigenvalues indicate significant mobility/flexibility. Here
is an idea for you to try : remove rotations/translations using not atoms
from the whole protein, but only the heavy atoms of the residues you are
interested in. You can even speed-up the calculation by generating a
DCD+PSF file containing only the selected residues (which will make carma
much faster).
The reason I am suggesting this is that the negative eigenvalues may come
from the fact that you look at a subset of atoms for which (subset)
overall rot/trans may still be present in your fitted trajectory.
> So you suggest when I have to make comparison between two sets of
> simulations its advisable to have same number of frames?
Your simulations (at ~90 ns) should be about 100K frames, which is not a
huge amount of frames. If the protein is not too big, I would rather use
all frames (ie. do not use '-step', just '-last').
My twocents,
Nicholas